pytdc

You can load AI-ready therapeutics datasets for drug discovery using ADME, DTI, and MolGen APIs provided by PyTDC. These modules offer standardized access to curated single-prediction, multi-prediction, and generation tasks for reproducible ML research.

A scaffold split partitions molecular datasets based on chemical structure scaffolds to evaluate model generalization. PyTDC provides built-in scaffold and cold splits to ensure robust testing of drug-discovery models on structurally unseen compounds.

Yes, PyTDC operates with Python and requires numpy and pandas as dependencies. You can process ADME, toxicity, and DTI datasets loaded via its APIs directly within your existing pandas and numpy data processing pipelines.

The best way to evaluate molecular generation models is using PyTDC's generation oracles and Evaluators. These utilities provide standardized benchmarking for goal-directed optimization workflows, enabling reproducible comparisons across different model outputs.

Yes, PyTDC provides benchmark group access and evaluation utilities to support reproducible therapeutics experiments. It enables standardized evaluation across multiple ADME, toxicity, DTI, and generation tasks without requiring manual data wrangling.

Oracles are needed to score and guide molecular generation workflows toward desired properties. PyTDC includes generation oracles that act as objective functions for goal-directed optimization, allowing models to be evaluated against standardized therapeutics benchmarks.